Evaluation method for stochasticity in seismic response of 3D sedimentary basins based on artificial neural network

The scattering of seismic waves by sedimentary basins leads to basin amplification effects and increases earthquake damage of structures. The geotechnical parameters of the basins generally have significantly stochastic characteristics, resulting in the variability of the basin response. This study aims to propose an efficient simulation method for ground motions of 3D sedimentary basins considering geotechnical uncertainty based on the fast multipole boundary element method (FM-IBEM) and artificial neural network (ANN). Firstly, an ANN model is constructed with geotechnical parameters, incident wave frequency and surface location as the input parameters and 3D sedimentary basin response as the output parameter. Secondly, a Monte Carlo simulation is utilized to evaluate the randomness of ground motions in sedimentary basins, and the ANN model and the existing results in the dataset are used instead of numerical simulations for sample solutions, which improves the computational efficiency of stochastic problems by reducing the computational time. The results indicate that the proposed method can be used to solve and evaluate the seismic response of 3D sedimentary basins considering the randomness of geotechnical parameters with high computational efficiency. The influences of randomness of geotechnical parameters on the seismic response of sedimentary basins are not negligible, and the variability of surface response of sedimentary basins is related to the location of surface points, incident frequency and distribution. It shows non-linear correlation characteristics with mean value distribution. The randomness of geotechnical parameters has a significant effect on ground motions of the basin under high-frequency incident waves. The coefficient of variation of the amplification effects can reach 0.3, which is three times the coefficient of variation of geotechnical parameters. The peak acceleration of the basin corresponding to 95% of confidence level increases by 56% compared to the mean value.